生成密集视差图的方法是LIBELAS（高效大规模立体视觉）。 此外，视频中还会显示将点云转换为不同的参考帧。 可视化显示在rviz中。 数据由安装在IIT Kharagpur的Jackal UGV机器人上的Point Gray Bumblebee立体相机收集。

Introduction

This article is a tutorial of the code implemented in the repository stereo-dense-reconstruction. The theory is not covered in detail, so some basic reading on the topic is suggested before one delves into the code. This ROS package computes dense disparity maps using LIBELAS and converts them into 3D point clouds. The method for generation of disparity maps can be changed based on user preferences. I had written an article on generating dense disparity maps using ORB descriptors earlier. That method can be implemented real time by using optimized code and parallelization.

链接：

http://sourishghosh.com/2017/dense-reconstruction-from-stereo-libelas/

代码链接：

https://github.com/sourishg/stereo-dense-reconstruction

原文链接：

http://weibo.com/5501429448/EC2HqpiMG?ref=home&rid=12_0_8_2669523763101168696&type=comment